In an effort to apply a large and realistic load, we connected the StealthXstream simultaneously to two platforms, similarly to the setup in our previous review. Making up a portion of the "required" additional load is a hungry Pentium Extreme 955 dual-core processor based in the inefficient NetBurst architecture, with a manually-set 1.45V core on both platforms further increasing load.

Even with all these parts, the AMD-based system draws only an estimated 265W at full load, and the Intel system draws 349W. Though the two combined will exceed 600W under full load, it's impossible to push all components to maximum power draw at the same time. Getting us as close as possible to maximum load is the benchmark 3DMark2001SE, which was run continuously in "looping mode" on both platforms simultaneously.

Systems with "power-hungrier" graphics cards will cause larger load changes than ours when starting games or simulations, so we substituted four ancient hard drives as load devices, connecting and disconnecting them while the system was running. This added dynamic load reaching approximately 50W maximum.

All the rails remained within allowable tolerance levels of plus or minus five percent, with the greatest amount of fluctuation seen on the "MB/Accessory" lead. The only voltage that initially seemed a bit sub-par is actually OK, the 5V rail at most 2.5 percent below its nominal value.

We didn't notice any noise from the StealthXstream, although that is in part because the CPU and graphics card coolers of the two attached platforms overwhelmed it. Ultimately, however, this should prove to be a similar case with more traditional builds in the real world.